Neuroinflammation and ER-stress are key mechanisms of acute bilirubin toxicity and hearing loss in a mouse model

Abstract

Hyperbilirubinemia (jaundice) is caused by raised levels of unconjugated bilirubin in the blood. When severe, susceptible brain regions including the cerebellum and auditory brainstem are damaged causing neurological sequelae such as ataxia, hearing loss and kernicterus. The mechanism(s) by which bilirubin exerts its toxic effect have not been completely understood to date. In this study we investigated the acute mechanisms by which bilirubin causes the neurotoxicity that contributes to hearing loss. We developed a novel mouse model that exhibits the neurological features seen in human Bilirubin-Induced Neurological Dysfunction (BIND) syndrome that we assessed with a behavioural score and auditory brainstem responses (ABR). Guided by initial experiments applying bilirubin to cultured cells in vitro, we performed whole genome gene expression measurements on mouse brain tissue (cerebellum and auditory brainstem) following bilirubin exposure to gain mechanistic insights into biochemical processes affected, and investigated further using immunoblotting. We then compared the gene changes induced by bilirubin to bacterial lipopolysaccharide (LPS), a well characterized inducer of neuroinflammation, to assess the degree of similarity between them. Finally, we examined the extent to which genetic perturbation of inflammation and both known and novel anti-inflammatory drugs could protect hearing from bilirubin-induced toxicity. The in vitro results indicated that bilirubin induces changes in gene expression consistent with endoplasmic reticulum (ER) stress and activation of the unfolded protein response (UPR). These gene changes were similar to the gene expression signature of thapsigargin-a known ER stress inducer. It also induced gene expression changes associated with inflammation and NF-κB activation. The in vivo model showed behavioural impairment and a raised auditory threshold. Whole genome gene expression analysis confirmed inflammation as a key mechanism of bilirubin neurotoxicity in the auditory pathway and shared gene expression hallmarks induced by exposure to bacterial lipopolysaccharide (LPS) a well-characterized inducer of neuroinflammation. Interestingly, bilirubin caused more severe damage to the auditory system than LPS in this model, but consistent with our hypothesis of neuroinflammation being a primary part of bilirubin toxicity, the hearing loss was protected by perturbing the inflammatory response. This was carried out genetically using lipocalin-2 (LCN2)-null mice, which is an inflammatory cytokine highly upregulated in response to bilirubin. Finally, we tested known and novel anti-inflammatory compounds (interfering with NF-κB and TNFα signalling), and also demonstrated protection of the auditory system from bilirubin toxicity. We have developed a novel, reversible, model for jaundice that shows movement impairment and auditory loss consistent with human symptoms. We used this model to establish ER-stress and inflammation as major contributors to bilirubin toxicity. Because of the rapid and reversible onset of toxicity in this novel model it represents a system to screen therapeutic compounds. We have demonstrated this by targeting inflammation genetically and with anti-inflammatory small molecules that offered protection against bilirubin toxicity. This also suggests that anti-inflammatory drugs could be of therapeutic use in hyperbilirubinemia.

Fig 1. Bilirubin exposure activates ER stress and NFκB pathways in SHSY5Y cells.

SHSY5Y cells were exposed to 50 μM bilirubin for 4h. Whole genome gene expression analysis was subsequently carried out. A) The gene expression pattern was compared to a compendium of 1600 chemical treatments using the cMap; the figure shows the correlation score for the top 10 chemicals, with bilirubin sharing greatest similarly to thapsigargin. B) The heatmap shows the correlation between the top 100 most changed genes in response to bilirubin or thapsigargin (n = 3). Samples and genes were clustered using the Euclidean distance algorithm. Up-regulated genes are shown in red and down-regulated genes in blue. The values represent log₂ fold change compared to control samples. C) Representative images of immunoblots measuring the protein expression levels for ER stress marker proteins and D) NF-ĸB marker proteins (n = 3). E) Densitometry-based quantification of the immunoblots where significant changes in protein expression were detected by ANOVA (*p<0.05, **p<0.01). F) Representative images of morphological changes in the ER caused by application of 50μM bilirubin. The ER was visualised using an antibody for the ER membrane protein, calreticulin. The scale bar represents 10 μm.

Fig 2. A mouse model for hyperbilirubinemia shows bilirubin toxicity on movement behavior and the auditory threshold.

A behavioural index (0–3, see methods) was used to assay bilirubin effects on movement and wellbeing over a 24h period in mice. A) Plot of the behavioural index from the time of IP injection in vivo across a range of bilirubin doses (in presence of 300 mg/kg sulfadimethoxine). Filled black circles: 550mg/kg (n = 3) for 0-4h after injection; filled red circles: 450mg/kg (n = 5 for 0-6h and n = 24 24h); Filled grey circles: 200mg/kg (n = 3); Open triangles: Control 0mg/kg (n = 3) were injected with sulfadimethoxine, alone. B) The dose-response for bilirubin (75-550mg/kg) on the mouse behavior assayed using the Behavioural Index, 4h after bilirubin injection. The mean ± SEM (n = 3 to 5 animals) is fit to the Hill equation giving a half-maximal toxicity of 341±22 mg/kg. C) Hearing function across the same time-course as bilirubin exposure was assayed by measuring the ABR: control traces (before bilirubin exposure) are shown plotted against time and with increasing sound intensity from 44–94 dB SPL. Blue dashed lines mark the peak ABR waves I-II-III. D) IP injection of 450 mg/kg of bilirubin caused a near complete loss of the ABR E) which recovered after 24 h. F). The ABR thresholds were measured at 4h or 24h after bilirubin injection (control is injection of sulfadimethoxine alone) = 40.1 ± 1.0 dB SPL, n = 14; bilirubin 4h = 91.5 ± 2.5 dB SPL, n = 4; bilirubin 24h = 38.2 ± 1.5 dB SPL, n = 9, ANOVA followed by Bonferroni P<0.001. Note that only one animal was tested at 4 and 24h, the other 8 animals were tested only at 24h, see methods).

Fig 3. Bilirubin toxicity induces ER stress and inflammation in a novel model of bilirubin toxicity.

A) Whole genome gene expression analysis was carried out on multiple brain regions following bilirubin exposure: brainstem, cerebellum, cochlear nucleus (CN) and the medial nucleus of the trapezoid body (MNTB). The top 100 genes with the greatest fold changes across all datasets are shown. Up-regulated genes are shown in red and down-regulated genes in blue. The values represent log₂ fold change compared to control samples. Samples and genes were clustered using the Euclidean distance algorithm. Quantification of the two mRNAs most highly up-regulated in response to bilirubin (mean±SEM, n = 4, arbitrary fluorescence units) for B) lipocalin-2 (LCN2) and C) S100A8. Samples marked with * are significantly changed (p<0.05) following a Welch t-test. Representative immunoblots for D) ER stress and E) NF-kB pathway marker proteins in the CN and MNTB following bilirubin exposure for 4 or 24h (n = 3). Where significant changes in protein expression were detected by ANOVA, densitometry-based quantification of the immunoblots are included (*p<0.05).

Fig 4. Lipopolysaccaride (LPS) a known neuro-inflammatory agent and bilirubin share significant similarities in their gene activation profile and both cause a rise in auditory threshold.

A) Whole genome gene expression analysis was carried out on brain tissue following bilirubin or LPS exposure. The top 100 genes with the greatest fold changes across all datasets are shown. Up-regulated genes are shown in red and down-regulated genes in blue. The values represent log₂ fold change compared to control samples (n = 4). Samples and genes were clustered using the Euclidean distance algorithm. B) The ABRs from a mouse are shown before LPS injection and C) 4h after LPS injection in the same animal. ABR responses are displayed over volume range of 35 to 90 dB (SPL), with the trace at threshold indicated in red. In this paired data LPS clearly raised auditory thresholds. D) Average data are plotted as Mean±SEM auditory thresholds, shown before and after LPS injection; LPS significantly increased auditory threshold (39±2 dB SPL for control n = 5 vs 50 dB ±7 dB SPL after LPS n = 4, p = 0.014, t test).

Fig 5. Suppression of neuroinflammation protects hearing from bilirubin toxicity.

ABR measurements were taken following bilirubin exposure combined with genetic and pharmacological intervention in inflammatory signalling. A-F) Representative ABR traces for each condition are plotted across an intensity range of 35-90dB, traces in red indicate the threshold, scale bar is 1μV. a) Control; b) Control exposed to bilirubin alone; c) Lipocalin-2 (LCN2KO) knockout mouse. d) Git 27 (TLR inhibitor); e) 3,6’Dtt (novel TNF-α inhibitor). f) Bay 11708 (NF-κB inhibitor). G) ABR threshold summary data with mean ± SEM; statistical significance was assessed by ANOVA (followed by Bonferroni, see text for numerical values). Comparisons are as indicated by the star * (p<0.001) with comparison of:- saline control vs bilirubin control; and bilirubin control and each test condition. H) Summary diagram illustrating the overall conclusion that multiple pro-inflammatory pathways interact to induce inflammation following bilirubin exposure.